jon-chuang
jon-chuang
Why? 319-bit prime modulus reduces G1 cost by approx 1.44x vis a vis BLS12-381/377. Constructing curves from Cocks-Pinch method (only known method to generate pairing-friendly curves of prescribed order) doubles...
We should standardise the use of the feature across all crates. This is required for macro based compilation in cuda kernels.
Not sure of lifetime of the src array, for instance, if it is donated. I can't run any tests locally as I have no CUDA device.
**System information** - Are you willing to contribute it (Yes/No): Maybe **Describe the new feature and the current behavior/state** Currently, we vendor modified tensorflow-alpa and use functorch repo. But functorch...
Triton `dot` out_dtype default is `float32` (see [here](https://github.com/openai/triton/blob/08c16589573621fcb8cd5a9c3b8a0537077f876d/python/triton/language/core.py#L988)) Currently, we do not lower any `out_dtype` However, perhaps it needs to be lowered based on `lax.dot_general(preferred_element_type=)`. Hence, performance cannot be tweaked...
There are at least 2 dependent projects now: - Intel's FastRAG which vendors the repo - llama_index: https://github.com/jerryjliu/llama_index/pull/6818 It would be nice if users could `pip install colbert-ir` instead of...
It is an interesting question whether ColBERT can be scaled up in one of the following ways: 1. Shard and distribute indexes with Ray actors 2. Utilize the sharded HNSW...
### Bug description When re-running autotune, the JIT system doesn't work properly Resulting in error: ``` JIT session error: Symbols not found: [ _runtime_llcl_Runtime___init__kA6A6AkA6A6A6A6AoApA, _runtime_llcl_AsyncTaskGroup_add_task_runtime_llcl_AsyncTaskGroup_builtin_coroutine_Coroutine_NonekA6A6AkA6A6A6A6AoAkA6A6AkA6A6AmAsAkA6A6AkA6A6AcBeBpA, _runtime_llcl_Runtime___exit___runtime_llcl_RuntimekA6A6AkA6A6A6A6AoAkA6A6AkA6A6ApA ] /__w/modular/modular/Kernels/mojo/builtin/_startup.mojo:76:1: error: no...
Hi, I am interested in your project. What is the current status? How long more till implementing a range proof/pedersen Hash? I would love to help in getting to the...
Hey guys awesome work. Quick question about k8s compat: If the kineto instrumented processes are running in K8s containers, are you guys aware if enabling host IPC, and running dynolog...